Skip to content

Latest commit

 

History

History
238 lines (215 loc) · 13.9 KB

CHANGELOG.md

File metadata and controls

238 lines (215 loc) · 13.9 KB

CHANGELOG

To find the version running on your machine use the command topicexplorer version. In older releases, this was vsm version or vsm --version.

Not all beta releases are documented below. We give the summary of all changes between major stable milestones. If you are using a beta version not listed below, we highly recommend switching to one listed below or simply updating using topicexplorer update.

This project follows the PEP 440 versioning conventions. See Semantic Versioning for justification. The only difference between PEP 440 and SemVer.org is the pre-release format 1.0b20 vs. 1.0-beta.20.

This CHANGELOG follows the conventions at Keep a CHANGELOG. Versions should group changes in the order: Added, Changed, Deprecated, Removed, Fixed, and Security (see section "What makes a good change log?").

Unreleased

1.0b182 - 2017-10-08

  • Added:
    • #200 Python 3 Support
    • Added support for the htrc-feature-reader.
    • Added topicexplorer export and topicexplorer import commands.
  • Changed:
    • #213 demo launches to browser
    • #210 isomap axis removed
    • #206 docs.json file very large, causes slow performance. Now defaults to 10 results
  • Fixed:
    • #218 Python 3: raw hex in prep frequency lists
    • #217 Python 3: init frequency filter b'string'
    • #216 demo/fulltext not working / Python 3
    • #215 ap.md not updating on Windows
    • #211 Bad link to Blei 2003 in AP demo
    • #208 UnicodeEncodeError when click on fulltext icon
    • #207 Python 3 error in topicexplorer update
    • #205 Mac OS Update has broken launch to browser (and Jupyter)
    • #204 ignore .DS_Store files on Mac
    • #203 Python 3 install_data error
    • #202 Unicode errors on combined 2+3 codebase during prep stage.
    • #194 Demo issue running within conda env

1.0b159 - 2017-04-26

  • Added:
    • Support for custom index.html in app.wsgi multi-model Apache config.
  • Changed:
    • Isomap now exmaines more nearest neighbors, which leads to lower numbers of overlapping topics in cluster view.
  • Fixed:
    • #199 warn user when term is not in corpus
    • #198 uBlock Origin blocks fingerprint.js from loading, preventing document search
    • #187 true button in handian document modals
    • #184 longer texts are not scrollable
    • #183 init is not selecting correct corpus structure when exclude files are present
    • Issue with decoding unicode URL parameters

1.0b146 - 2016-11-22

  • Added:
    • #144 permissions string - modal now accessible via © button in lower left.
    • Sidebar is now on all subpages.
    • Topic Fingerprint view now available via the Document view page.
  • Changed:
    • All pages now load from a single master.mustache.html filled in with the appropriate subpage.
    • Upgraded all sites to Bootstrap 3.
  • Fixed:
    • topicexplorer demo now has a corpus description.
    • #177 1.04b145 unpickling error
    • #172 unicode error with metadata import
    • #171 term search for topics internal server error
    • #169 directory structure error during init
    • #104 Metadata import
    • Variety of rendering errors in the Document-view page.
    • Fixed issue with histogram bar width

1.0b134 - 2016-10-09

  • Added:
    • documentation for BibTeX support.
    • #153 Add default index.html
    • #143 Links to corpus home page
    • #99 Add -q to all commands to run without prompts
  • Changed:
    • #154 Modal fulltext default
    • #120 "continue training" is ambiguous
    • Updated vsm to 0.4.0b6
  • Fixed:
    • Metadata now keeps all information in unicode.
    • topicexplorer metadata now allows for label updates with --rename flag.
    • Upgrade to D3 for topic cluster display broke other parts of visualization. Reverted D3 and removed chargeDistance property from cluster view.
    • #168 Corpus rebuild prompt does not work correctly.
    • #166 Invalid attribute 'quiet'?
    • #165 Continue training does not remove old clusters
    • #163 Continue training does not update model_pattern

1.0b121 - 2016-09-18

  • Added:
    • topic cluster display
  • Fixed:
    • #160 - New isomap display doesn't load in Safari
    • Page labels in HTRC extension
    • BibTeX extension config file loading
    • topicexplorer metadata now works with all context types

1.0b111 - 2016-08-14

  • Added:
    • topicexplorer metadata command with list, export, and import
  • Changed:
    • Ran autopep8 on topicexplorer, topicexplorer.extensions and topicexplorer.lib
  • Removed:
    • Dependency for pyenchant.
  • Fixed:
    • Duplicate topics (fixed via vsm==0.4.0b4 regression)
    • Error in topic view due to uncast float.

1.0b107 - 2016-07-30

  • Added:
    • mod_wsgi support
  • Removed:
    • Module topicexplorer.launch. Alias for command topicexplorer launch still works and is documented way to launch browser. Reserve topicexplorer serve for advanced use.
  • Fixed:
    • #37 - Merge ports to single in-page model switch

1.0b106 - 2016-07-25

  • Added:
    • New tutorial notebook.
    • Mendeley support via BibTeX export (undocumented).
    • Chinese language support using pymmseg segmenter:
      • Modern Chinese (Mandarin): topicexplorer init --tokenizer zh
      • Classical Chinese: topicexplorer init --tokenizer ltc
      • Classical Chinese (alias): topicexplorer init --tokenizer och
    • Support for hard-coded phrasal tokenizers, should inspire generic SKOS tokenizers:
  • Changed:
    • Massive load performance increases from upstream vsm==0.4.0b1.
    • #136 - Rename master command from vsm to topicexplorer
  • Deprecated:
    • vsm commands will be removed in v1.0.
  • Removed:
    • Developer version of vsm from dependency_links in setup.py.
  • Fixed:
    • #116 - "vsm serve" should launch the browser - consistent with notebook and launch
    • #122 - Number of seeds must be equal to number of processors
    • #123 - update documentation to show PowerShell for windows
    • #127 - typo in SEP url
    • #128 - ini file is created relative to modeled dir, not pwd
    • #132 - Handian corpus is saying "similarity of documents in the HandianCorpus to None"
    • #134 - unicode being decoded in search box

1.0b88 - 2016-04-29

  • Added:
    • Support for sentence-level modeling: vsm init --sentences
    • Default stop frequency argument: vsm init --freq 5
    • Quiet argument and nulls in vsm prep: vsm prep -q
    • Multi-process PDF conversion using concurrent.futures in vsm init.
  • Changed:
    • Improve train memory performance from upstream changes in vsm==0.4.0a26
    • Improve prep memory performance from upstream changes in vsm==0.4.0a23
  • Fixed:
    • Error in vsm demo (upstream)
    • Error in vsm prep with null values
    • Error in vsm langspace

1.0b79 - 2016-04-13

  • Added:

    • Very basic support for Chinese word segmentation: vsm langspace
    • #94 - Create generic fulltext file serving for pdfs and text files
    • #97 - Add langdetect to vsm prep to use guesses for stopwords.
      • This change was reverted in 0048d60 due to performance issues.
  • Changed:

    • Improve init and prep performance from upstream changes in vsm==0.4.0a20
    • No longer perform sentence mining by default. Significant speedup to vsm init for most use cases.
    • label_module semantics now use init(viewer, config, args)
    • #113 - Topic view should default to normalized
  • Fixed:

    • Fixed dynamic port switching
    • Fixed plus signs in doc_ids.
    • Fixed popover/search box z-index issue.
    • #93 - continuation of training with fewer words / topic explorer can hang
    • #107 - AP Demo text opening error
    • #108 - Capture missing nltk libraries and suggest downloader command
    • #111 - explorer error on TJBooks corpus
    • #114 - vsm update on Windows fails clumsily when vsm.exe is running
    • #115 - vsm launch should fail more gracefully.

1.0b60 - 2016-02-20

  • Added:
    • Native benchmarking tools via vsm -p and vsm -t
    • ASCII histograms for vsm prep high and low frequency filters.
    • License badge for README.md
    • #98 - Add --dry-run to help with automation in vsm train
  • Changed:
    • Improve init performance from upstream changes in vsm==0.4.0a11
    • Migrated vsm --version to vsm version
    • #105 - vsm init without .ini extension should auto-suggest
  • Removed:
    • NP-hard graph coloring algorithm removed from topic explorer launch.
  • Deprecated:
    • vsm --version will be removed in a future version.
  • Fixed:
    • #83 - "Close" button not appearing in reduced window
    • #100 - Windows vsm update does not run
    • #101 - Check if repo is on master branch in vsm update
    • #102 - entering 0 for minimum word occurrence rate during vsm prep causes error

1.0b41 - 2016-02-02

  • Added:
    • vsm update command.
    • vsm --version command.
    • release.py automation.
    • Notes in README.md on Bug Reports
  • Changed:
    • Import upstream fixes via vsm==0.4.0a8
  • Fixed:
    • vsm init --unicode fixed.
    • #82 - Alphabetization not working
    • #86 - Capitalization inconsistencies
    • #89 - color map bug affecting browser launch

1.0b32 - 2016-01-20

  • Start of CHANGELOG.md
  • Added:
    • PDF file support
    • --unicode/--decode flags.
    • Progress bars
  • Changed:
    • vsm prep now uses a single stoplist pass, using an in-place rather than out-of-place technique. Massive performance increase.